Skip to content

device-panther 0.1.1: the GPU, the haptics and suspend-to-idle - #145

Merged
derekross merged 2 commits into
derekross:mainfrom
nesmeck:panther-gpu-haptics
Oct 30, 2026
Merged

derekross merged 2 commits into
derekross:mainfrom
nesmeck:panther-gpu-haptics

Conversation

@nesmeck

@nesmeck nesmeck commented Oct 29, 2026 •

Copy link
Copy Markdown
Contributor

omamobile-device-panther 0.1.1, plus extract-vendor-firmware -d panther and the notes from the Pixel 7's GPU and haptics boots. The kernel side is derekross/linux-omarchy-mobile#129 (GPU) and derekross/linux-omarchy-mobile#130 (haptics, on top of #129).

Package. Three files from the Pixel 7 Pro's package, byte-identical; the package test compares each with cheetah's:

The two udev rules match nothing until a kernel has the GPU or haptics node, so the package is safe with today's omamobile.

tools/extract-vendor-firmware -d panther. It extracts the same gs201 vendor files under the same names as cheetah, with panther's board and codename substituted, and checks the haptics and media deltas as it does for cheetah. Run on this phone's factory image (CP3A.261005.002.A1), it exits 0 and gives cs40l26.wmfw and cs40l26.bin; that is what the haptics test used.

Notes (device/panther/NOTES.md, 2026-10-28):

  • the GPU boot: panthor on the Mali-G710, and Hyprland rendering on it through the project's Mesa;
  • the haptics boot: firmware 7.2.91 with the wavetable, and ff-rumble effects felt;
  • power off restarts the phone. A shutdown hook saved the kernel log to userdata ten times a second, then powered off itself with sysrq o. Nothing was saved after the trigger, so the reset comes within the first tenth of a second of kernel_power_off(), from PSCI SYSTEM_OFF (syscon-poweroff was disabled in the test DT). The Pixel 7 Pro hasn't powered off from Linux either (SUSPEND-PLAN P7), so this may be useful there;
  • USB after a replug: the MAX77759 data switch stays open, as on the other Tensor phones. omamobile-usb-rescue needs USI13, which the panther board file doesn't enable yet; that's the next kernel step.

Matrix. The Pixel 7 row and BRINGUP-M1 now say #108 and #109 are merged:

The rest of #79's follow-ups (Touch "works", the touch drop-ins, CONFIG_TOUCHSCREEN_FOCALTECH_FT3658=m) come with the touch PR.

Checks: tests/all (194 ok, on main at a6034a5), tools/pr-check upstream/main, and the panther package's test, both from the repository and from a lone copy of the package directory.

Phone evidence (CONTRIBUTING §6.5). The phone runs this package, 0.1.1, built and installed on it with tools/phone-install-package (its check() ran test/all on the lone package directory and passed), with the hand-installed rule copies removed. The kernel is derekross/linux-omarchy-mobile#134 at ace1ba57f merged with derekross/linux-omarchy-mobile#133 at fdc4f4403, the build T6 tested (g8d46192a301e), plus one local commit on top that enables the MAX77759's I2C bus (USI13) for omamobile-usb-rescue, the next kernel PR.

phone-smoke: panther

Kernel 7.3.0-rc6-omamobile-push1-g9ec3a2e38705, tools at 6177949, 2026-10-29T20:57Z.

Check Result Detail
boot (required) ok kernel 7.3.0-rc6-omamobile-push1-g9ec3a2e38705, slot b, up 28892s
storage guard (required) ok 55 partitions; only userdata and slot B's boot images writable
root (required) ok ext4 on userdata, 99G free
kernel health (required) ok no oops or BUG; 0 WARN, 0 error lines in dmesg
systemd units ok none failed
USB gadget (required) ok bound to 11210000.usb
display ok card0-Unknown-1 connected, 1080x2400
GPU ok render node on panthor
session (required) ok Hyprland and the Omarchy shell running
lock ok PAM file present, the shell can lock (passwordPam true, locked true)
touch ok FocalTech FT3658
buttons ok gpio-keys
Wi-Fi skip no wireless interface
Bluetooth skip no controller
battery skip no battery power supply
sound cards skip none
sensors skip no sensor proxy
cameras skip no libcamera tools

Packages: hyprland 0.56.2-5.1, libcamera 0.7.2-4.21, mesa 1:26.2.4-1.4, omacam 0.3.3-1, omamobile-apps 0.2.0-1, omamobile-device-panther 0.1.1-1, omamobile-firstboot 0.1.2-1, omamobile-keyring 20261026-1, omamobile-mirror 0.1.1-1, omamobile-omarchy 0.1.12-1, omamobile-owl-widgets 0.1.1-1, omamobile-shell 0.2.28-1, omarchy 4.0.6-1, quickshell 0.3.2.r7.g2b4501a-1

🤖 Generated with Claude Code

@nesmeck

nesmeck commented Oct 29, 2026

Copy link
Copy Markdown
Contributor Author

The failing Tests job is in omamobile-audio, which this PR doesn't touch: "watchdog: in stage c9 a 15 C jump of the estimate mutes" (packages/omamobile-audio/test/all:457). It looks like a timing race on the runner: the test raises the estimate from a background sleep 0.4 while the watchdog watches for a fixed window, and here the watchdog's log shows it started but never saw the jump. The same test passes on main at the same base (40d132a) and three runs out of three locally. Could you re-run the job when you get a chance? I can't re-run it from the fork.

🤖 Generated with Claude Code

@W4HO-ham

Copy link
Copy Markdown
Contributor

That flaky stage c9 jump test is fixed in #150: it now waits for the watchdog's first reading, as the other jump tests already do.

@derekross

Copy link
Copy Markdown
Owner

Thanks, Alejandro. This is a tidy follow-up to #79. The three files are byte-identical to the Pixel 7 Pro's and the package test compares them, so they can't drift. The s2idle drop-in closes the gap I was worried about while the board file has no mem_sleep_default. And extract-vendor-firmware -d panther keeps cheetah's table exactly as it was: I extracted the map for -d cheetah at main and at your head and the two are identical. Panther's map also gets the same format checks, so an image that doesn't match fails rather than staging the wrong files. The power-off investigation (a shutdown hook logging to userdata ten times a second) is a clever measurement, and useful for the Pixel 7 Pro's P7 too.

Two things before I merge:

  1. The package test fails when the package is built on its own. The new loop that cmps the two udev rules against cheetah's copies (test/all:78) sits outside the if [[ -f ../omamobile-device-cheetah/src/reboot-to.c ]] block that guards the other comparisons. tools/phone-install-package sends only the package directory to the phone, and makepkg's check() runs test/all there. tools/build-packages --container-check does the same in its container. In both places ../omamobile-device-cheetah doesn't exist. I copied the package directory alone into a temp directory and ran its test:

    ok - scripts parse
    ok - build and install stay inside the package directory
    ok - pacman gets the keyring, the mirrorlist, and the install script
    ok - reboot-to compiles
    not ok - 20-omamobile-gpu-always-on.rules matches the Pixel 7 Pro's (the same GPU and haptics)
    

    Main's 0.1.0 passes the same run. Could you move the cmp loop into the guarded block, with a pass line like its neighbours? Please also add a content check outside the guard, so a lone copy still tests something. For example: grep -q 'KERNEL=="28000000.gpu", ATTR{power/control}="on"' on the staged GPU rule, and grep -q 'ATTRS{name}=="cs40l26_vibra"' on the feedbackd one, the way the sleep drop-in already gets its MemorySleepMode=s2idle grep.

  2. The phone evidence in the description. This changes a package, so CONTRIBUTING §6.5 and the template ask for a tools/phone-smoke report from the phone running it, with uname -r. A report from the T6 boot in device/panther: the touchscreen (FT3658 kernel option, T6) and the touch notes #149 (or any later boot with 0.1.1 installed) is fine. Please also add the kernel release to the 2026-10-28 GPU and haptics entries in NOTES.md, so each result is tied to a build.

The red Tests job is the audio package's timing race, as you said: #149 carries this same commit and passed on the same base, and #150 fixes the test. I'll re-run it after your push.

One more thing since you pushed: I merged #140 (a DEVICES.md and panther-notes refresh after #108), and it touches the same panther lines, so this branch now conflicts with main in device/panther/BRINGUP-M1.md, device/panther/NOTES.md and docs/DEVICES.md. Please rebase onto main when you push the test fix. Your text is newer than #140's, so take your side wherever the two disagree. Sorry for the churn.

On order: this one merges first, and #149 after it, rebased. I'll explain the details there.

derekross pushed a commit that referenced this pull request Oct 29, 2026
…st reading (#150)

The stage c9 jump test still raised the estimate 0.4 s after starting the watchdog, so on a slow runner the watchdog took the raised value as its baseline and never muted (seen on #145). It now waits for the watchdog's start line, as the other jump tests do.
nesmeck and others added 2 commits October 29, 2026 13:40
The Pixel 7 Pro's GPU always-on and feedbackd udev rules, and its
sleep.conf.d drop-in, byte-identical (the package test compares them).
mem_sleep reads "s2idle [deep]" on these kernels, and gs201 never wakes
from deep, so Suspend has to be s2idle before anyone taps it.

tools/extract-vendor-firmware takes -d panther: the same gs201 vendor
files under the same names as cheetah, with panther's board and codename,
so the CS40L26 gets this phone's own cs40l26.wmfw and cs40l26.bin.

NOTES records the GPU and haptics boots (kernel derekross#129 and derekross#130), that
power off restarts the phone within the kernel's first tenth of a second
(PSCI SYSTEM_OFF, measured with a shutdown hook that logged to userdata),
and that USB stays off the bus after a replug until USI13 is enabled for
omamobile-usb-rescue. The Pixel 7 row and BRINGUP-M1 say derekross#108 and derekross#109
are merged, and link the new kernel PRs from the GPU and Haptics cells.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…1/H1 kernels

The cmp of the GPU and feedbackd rules against cheetah's copies sat
outside the guard that the other comparisons use, so test/all failed
wherever the package directory travels alone (phone-install-package,
build-packages --container-check). Move it inside the guard, with a pass
line, and check the staged rules' content outside it, as the sleep
drop-in already is.

NOTES' 2026-10-28 GPU and haptics entries now name the kernel release
each result came from.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@nesmeck
nesmeck force-pushed the panther-gpu-haptics branch from a69b7d1 to 6177949 Compare October 29, 2026 21:03
@nesmeck

nesmeck commented Oct 29, 2026

Copy link
Copy Markdown
Contributor Author

Thanks for the careful review. Pushed (6177949), rebased onto main at a6034a5:

  1. The lone-package test. The cmp of the two rules is now inside the guarded block, with its own pass line ("the GPU and feedbackd rules match the cheetah package"). Outside it, the staged rules get content checks: KERNEL=="28000000.gpu", ATTR{power/control}="on" on the GPU rule and ATTRS{name}=="cs40l26_vibra" on the feedbackd one. A copy of the package directory alone passes test/all, and so did makepkg's check() on the phone through tools/phone-install-package.
  2. Phone evidence. A tools/phone-smoke report is in the description: 0.1.1 installed with phone-install-package, every check ok, on omamobile-omarchy 0.1.20: the sleep lock turns the display off once the session is secure #134 + device/lynx: WIFI-PLAN after the fs_ready boot #133 (the T6 build) plus one local commit that enables USI13, as described there. NOTES' 2026-10-28 G1 and H1 entries now name their kernels (…-ga92e2bd8546c and …-g575bb567ca5f) and the local commits they were built from.
  3. The rebase. BRINGUP-M1.md, NOTES.md and docs/DEVICES.md each had one conflict with DEVICES: status after the 2026-10-28 evening merges #140, and I took this branch's side in each hunk only, so the raven and tegu rows from main are kept. The panther files are byte-identical to the previous head.

One thing I hit on the way that may be worth a look in the tooling: pacman -U in phone-install-package failed with omamobile: key "4185…F89F" is unknown, then database 'omamobile' is not valid. That fingerprint is my contributor key (keys/panther): the phone's omamobile.db came from my local build during install-phone, but the key isn't in the phone's keyring (omamobile-keyring 20261026 has only the release key). Since pacman -U checks every sync database, a contributor phone can't install a locally built package once its own database is in place. I didn't import the key. I installed this one with a temporary copy of pacman.conf without [omamobile] and left the real one untouched. It's probably the same thing as the phone-smoke key prompt you mentioned.

🤖 Generated with Claude Code

@derekross
derekross merged commit 47554d5 into derekross:main Oct 30, 2026
2 checks passed
@derekross

Copy link
Copy Markdown
Owner

Thanks, Alejandro. All three points are done, and done well.

  • The test. I ran test/all from the repository (13 ok) and from a copy of the package directory alone (5 ok, rc 0). Then I broke things on purpose. A changed cheetah copy of the feedbackd rule fails the in-tree comparison. A lone copy whose feedbackd rule no longer names cs40l26_vibra fails the new content check. So both halves now test what they claim.
  • The evidence. The smoke report from a phone running 0.1.1 is what I asked for, and so are the kernel releases and local commits on the G1 and H1 entries. Thank you for being clear about the extra USI13 commit in the smoke kernel.
  • The rebase. It's clean against today's main, and the raven and tegu rows from DEVICES: status after the 2026-10-28 evening merges #140 are intact. CI is green on 6177949, and the policy check passes.

I'm merging this now. The GPU node (derekross/linux-omarchy-mobile#129) can follow it, since the always-on rule is in the package.

One small thing for #149's rebase, and my fault for telling you to take your side everywhere: BRINGUP-M1's status line now says the FT3658 "had no mainline driver". #140 had changed that to say the chip never answered in boots 7 and 8 because of the supplies, which is what the body of that file and TOUCH-PLAN record (boots 6 and 7 had the driver). #149 touches the same lines. When you rebase it, could you put that sentence back, or better, point it at TOUCH-PLAN's result?

Thanks also for the pacman -U report. It's a real gap: a contributor phone whose [omamobile] database is signed with a local key can't verify that database, so every pacman -U stops. You were right not to import the key into the phone's keyring. I'll open an issue for phone-install-package and the local-key flow so it doesn't get lost here.

derekross added a commit that referenced this pull request Oct 30, 2026
The matrix and the Wanted list against what has merged since #140, with
each pull request's state checked on GitHub today:

- Pixel 7 (panther): the GPU and haptics nodes (kernel #129, #130) were
  closed because they landed unchanged through kernel #134, merged with
  the touchscreen node and s2idle by default; device-panther 0.1.1 (#145)
  carries the GPU's always-on rule, the feedbackd rule and the sleep.conf
  drop-in. GPU and Haptics stay "testing" and now cite #134, #145 and the
  2026-10-28 notes. Touch stays WIP: kernel #134's body records touches
  driving Hyprland and the PIN screen, but the driver fixes (kernel #133)
  and the kernel option with the touch notes (#149) are still in review.
  Suspend: "no" with the s2idle default named, since the notes say it
  hasn't run.
- Pixel 6 Pro (raven): Haptics "no" -> "works": HAPTICS-TORCH-PLAN H5
  passed (2026-10-28/29), cs40l25-vibra and the node merged as kernel
  #114, feedbackd's rule in device-raven 0.1.27 (#113). Display names
  its open kernel PR, #116.
- Pixel 7a (lynx): Bluetooth "no" -> WIP: the merged WIFI-PLAN (#133)
  records it scanning through its own UART; kernel #98 and #98 are in
  review. The Wanted line says so too.
- Pixel 7 Pro (cheetah) detail: Boot names the AVB footers a chained
  vbmeta_b needs (NOTES 2026-10-28 and 10-29, #104); Suspend describes
  the merged tier 1 auto-sleep (#64, off by default behind the
  omamobile-sleep toggle, P0-P5 on 2026-10-26) in place of the old
  "until S8" text. Nothing from the open #171, #172, #174-#176.
- Wanted: coral's #52 is merged, so its line no longer says "in review";
  flame gets a line for its native install in review (#55, #56, #58,
  #59, split out of #17); the OnePlus 13 (dodge) is listed with #111 and
  kernel #115, both open.

The Pixel 9a row and the Pixel 9 family line are left to #165. Every
status word still parses with the website's parseCell.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…st reading (derekross#150)

The stage c9 jump test still raised the estimate 0.4 s after starting the watchdog, so on a slow runner the watchdog took the raised value as its baseline and never muted (seen on derekross#145). It now waits for the watchdog's start line, as the other jump tests do.
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…ross#145)

The Pixel 7 gets three files from the Pixel 7 Pro's package, byte-identical
and compared by the package test when cheetah's package is beside it:

- sleep.conf.d/10-omamobile.conf: Suspend means s2idle. The phone offers
  "deep" (mem_sleep reads "s2idle [deep]"), and gs201 never wakes from it.
- 20-omamobile-gpu-always-on.rules, for Panthor on the Mali-G710
  (derekross/linux-omarchy-mobile#129).
- 72-omamobile-feedbackd.rules, for the CS40L26 haptics
  (derekross/linux-omarchy-mobile#130).

The udev rules match nothing until a kernel has those nodes. A lone copy
of the package (phone-install-package, build-packages --container-check)
checks the rules' content instead of comparing them.

tools/extract-vendor-firmware -d panther extracts the same gs201 files as
cheetah under panther's board name, with the same format checks; cheetah's
map is unchanged.

device/panther/NOTES.md records the GPU (G1) and haptics (H1) boots with
their kernel releases, power off restarting the phone (PSCI SYSTEM_OFF),
and USB not coming back after a replug. DEVICES.md and BRINGUP-M1 follow.

Tested on the Pixel 7 with 0.1.1 installed through phone-install-package:
tools/phone-smoke all ok on 7.3.0-rc6-omamobile-push1-g9ec3a2e38705.
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
The matrix and the Wanted list against what has merged since derekross#140, with
each pull request's state checked on GitHub today:

- Pixel 7 (panther): the GPU and haptics nodes (kernel derekross#129, derekross#130) were
  closed because they landed unchanged through kernel derekross#134, merged with
  the touchscreen node and s2idle by default; device-panther 0.1.1 (derekross#145)
  carries the GPU's always-on rule, the feedbackd rule and the sleep.conf
  drop-in. GPU and Haptics stay "testing" and now cite derekross#134, derekross#145 and the
  2026-10-28 notes. Touch stays WIP: kernel derekross#134's body records touches
  driving Hyprland and the PIN screen, but the driver fixes (kernel derekross#133)
  and the kernel option with the touch notes (derekross#149) are still in review.
  Suspend: "no" with the s2idle default named, since the notes say it
  hasn't run.
- Pixel 6 Pro (raven): Haptics "no" -> "works": HAPTICS-TORCH-PLAN H5
  passed (2026-10-28/29), cs40l25-vibra and the node merged as kernel
  derekross#114, feedbackd's rule in device-raven 0.1.27 (derekross#113). Display names
  its open kernel PR, derekross#116.
- Pixel 7a (lynx): Bluetooth "no" -> WIP: the merged WIFI-PLAN (derekross#133)
  records it scanning through its own UART; kernel derekross#98 and derekross#98 are in
  review. The Wanted line says so too.
- Pixel 7 Pro (cheetah) detail: Boot names the AVB footers a chained
  vbmeta_b needs (NOTES 2026-10-28 and 10-29, derekross#104); Suspend describes
  the merged tier 1 auto-sleep (derekross#64, off by default behind the
  omamobile-sleep toggle, P0-P5 on 2026-10-26) in place of the old
  "until S8" text. Nothing from the open derekross#171, derekross#172, derekross#174-derekross#176.
- Wanted: coral's derekross#52 is merged, so its line no longer says "in review";
  flame gets a line for its native install in review (derekross#55, derekross#56, derekross#58,
  derekross#59, split out of derekross#17); the OnePlus 13 (dodge) is listed with derekross#111 and
  kernel derekross#115, both open.

The Pixel 9a row and the Pixel 9 family line are left to derekross#165. Every
status word still parses with the website's parseCell.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants